Download Identification of Two Maternal Transmission

OPEN
SUBJECT AREAS:
GENOME-WIDE
ASSOCIATION STUDIES
GENETIC LINKAGE STUDY
HAPLOTYPES
Identification of Two Maternal
Transmission Ratio Distortion Loci in
Pedigrees of the Framingham Heart
Study
RARE VARIANTS
Yang Liu1,4, Liangliang Zhang1,4, Shuhua Xu2, Landian Hu1,4, Laurence D. Hurst3 & Xiangyin Kong1,4
Received
4 January 2013
Accepted
17 June 2013
Published
5 July 2013
Correspondence and
requests for materials
should be addressed to
X.Y.K. (xykong@sibs.
ac.cn)
1
The Key Laboratory of Stem Cell Biology, Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy
of Sciences, 2Max-Planck Independent Research Group on Population Genomics, CAS-MPG Partner Institute for Computational
Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200031, China, 3Department of
Biology and Biochemistry, University of Bath, Bath, UK, 4Ruijin Hospital, Shanghai Jiaotong University School of Medicine, Shanghai
200025, China.
Transmission ratio distortion (TRD) is indicated by the recovery of alleles in offspring in non-Mendelian
proportions. An assumption of Mendelian proportion is central to many methods to identify
disease-associated markers. This seems reasonable as, while TRD cases have been occasionally observed in
various species few instances have been identified in humans. Here we search for evidence of paternal or
maternal TRD with genome-wide SNP data of pedigrees from the Framingham Heart Study. After excluding
many examples as better explained by genotyping errors we identified two maternal-specific TRD loci for
autosomal SNPs rs6733122 and rs926716 (corrected P 5 0.029 and P 5 0.018) on LRP2 and ZNF133,
respectively. The transmission ratios were as high as 1.7,1.851. Genotyping validation and further
replication is still necessary to confirm the TRD. This study shows that there may be large-effect
maternal-specific TRD loci of common SNPs in the human genome but that these are rare.
T
ransmission ratio distortion (TRD) is indicted when the transmission ratios between different alleles deviate
from the 151 ratio that is expected under Mendelian inheritance. TRD has been observed and studied in
some model organisms, such as the segregation distorter system in Drosophila1 and the t-complex system in
mouse2. TRD is central to understanding the fate of alleles and in the longer term TRD associated loci may be
important for processes such a hybrid sterility3. A more immediate concern is that some genetic mapping
approaches, such as sib-pair linkage analysis and transmission disequilibrium test (TDT), can be affected by
TRD, since these approaches assume Mendelian inheritance4.
A few TRD loci have been identified in humans to date, and most of them were observed on rare variants such
as the Robertsonian translocations, which were often directly associated with human disease5. Compared to these
rare variants, TRDs on high-frequency variants such as SNPs can have more profound effects on human
populations, but very few of common-variant TRD has been identified. A previous study suggested that smalleffect TRDs may be very common throughout the human genome, while no large-effect TRD locus was identified6. Another study searched for general TRD loci with genome-wide SNP data; however, they found that most
of the identified loci were false positives due to genotyping errors and it was hard to distinguish true TRD loci
from the results7. In this study, we specifically searched for loci showing either paternal or maternal TRD with
genome-wide SNP data. We robustly identify two large-effect maternal-specific TRD loci of SNPs in humans.
Results
We used the large-scale genome-wide SNP data of individuals from the Framingham Heart Study (FHS) pedigrees (see Methods) to identify loci subject to paternal-specific and maternal-specific TRD. After quality-control
of the data and statistical tests (see Methods), a total of 41 paternal and/or maternal TRD SNPs reached the 5%
significance threshold after Bonferroni correction, equivalent to a raw p-value of 7.2 3 1028 (Supplementary
Table S1). After visual inspection of the cluster plots of the 41 SNPs, we found that 39 of them showed very poor
genotyping quality, and the corresponding TRDs were probably caused by the misclassification of genotypes
SCIENTIFIC REPORTS | 3 : 2147 | DOI: 10.1038/srep02147
1
Sex- differentiated P
5.3 3 1024
3.1 3 1023
4.2 3 10 (0.029)
2.6 3 1028 (0.018)
The SNP positions are in NCBI build 36. Chr: chromosome. MAF: minor allele frequency. HWE: Hardy-Weinberg Equilibrium.
G/A (0.23)
G/C (0.37)
2
20
rs6733122
rs926716
169707961
18221632
LRP2
ZNF133
0.89
0.47
182/170
212/241
228/125 (64.6%)
176/297 (37.2%)
0.52
0.17
28
Gene
Chr
Position
Major/minor
allele (MAF)
SNP
Table 1 | Two identified maternal TRD loci
HWE P
Maternal major/minor
Paternal major/minor allele transmissions (major
allele transmissions
allele transmission %)
Paternal
TRD P
Maternal TRD
P (corrected P)
www.nature.com/scientificreports
SCIENTIFIC REPORTS | 3 : 2147 | DOI: 10.1038/srep02147
(Supplementary Fig. S1). The genotyping errors affected both paternal and maternal transmission ratios of almost all autosomal SNPs in
the 39 SNPs (all paternal and maternal TRD P , 0.05 except for
rs17628931, which gave a maternal TRD P 5 0.056).
One of the remaining two SNPs, rs6733122, showed slightly mixed
clusters of the heterozygotes and the major homozygotes (Supplementary Fig. S1). Therefore, we checked the cluster plot of its most
strongly linked SNP, rs2284681 (r250.98). This SNP showed a good
clustering (Supplementary Fig. S1); however, it has a less significant
TRD effect (maternal TRD P 5 2.2 3 1027, corrected P 5 0.15).
Therefore, it is still possible that the TRD of rs6733122 was affected
by genotyping artifact that yielded more transmissions for the major
allele, and additional efforts in genotyping for this SNP would be
helpful to confirm the TRD. The other remaining SNP rs926716
showed a good clustering (Supplementary Fig. S1).
The remaining two SNPs gave genome-wide-significant maternal
TRD p-values of corrected P 5 0.029 and P 5 0.018 for rs6733122
and rs926716, respectively (Table 1). Interestingly, the paternal TRD
p-values were not significant (P . 0.1), and we observed significant
sex-differentiated TRD effects for the two identified SNPs (Table 1).
In addition, assuming that the paternal TRD effects of these two
SNPs existed and the transmission ratios were as small as
55%545%, there was still 88% and 77% statistical power to detect
the small paternal TRD effects of rs6733122 and rs926716, respectively, according to the total informative paternal transmission counts
(352 for rs6733122 and 453 for rs926716) and the observed paternal
TRD p-values (0.52 and 0.17). Therefore, the paternal TRD effects
should be minimal and we considered that the two identified TRD
loci were maternal-specific.
To avoid the situation that the TRDs were generated by a few
pedigrees and to evaluate the robustness of the TRDs, we performed
pedigree-based cross validation for the two maternal TRD loci (see
Methods). The overall cross validation consistencies in the 100 repetitions of ten-fold cross validation were 953 and 962 (out of 1000) for
rs6733122 and rs926716, respectively.
It is possible that the identified TRDs were confounded by actual
phenotypic association if the studied individuals, especially offspring, were sampled with enrichment of certain disease (e.g., for
TDT) or extreme trait, rather than randomly sampled from the
population. We confirmed that the individuals of the FHS were
sampled in a random way8–10. To avoid confounding by possible
biased sampling towards the phenotypes of interest in the
Framingham Heart Study, we also performed population-based
and family-based association analyses for the two identified TRD
loci with ten available phenotypes in the data (see Methods). Eight
of the phenotypes are quantitative traits; namely, height, weight,
systolic blood pressure, diastolic blood pressure, cholesterol level,
high-density lipoprotein level, triglycerides level and blood sugar
level, and the other two phenotypes are two disease statuses of diabetes and hard coronary heart disease. No significant association was
identified (all corrected P . 0.05). Furthermore, we used the Genetic
Association Database11 and GWAS Catalog12 to search for known
disease-associated loci within the two 100-kilobase surrounding
regions centered by rs6733122 and rs926716, and no disease or trait
association was found.
The first maternal TRD locus that we identified, rs6733122, is
located at the 39-end of a large gene LRP2 (Fig. 1a), and the corresponding LD block of rs6733122 extended for 3 kb (Fig. 1b, plotted
with Haploview13). The effect of the maternal TRD was quite large, as
the transmission percentage of the major allele G was 64.6%
(Table 1). LRP2, which encodes the megalin, is dominantly expressed
in thyroid and moderately expressed in kidney (BioGPS14). Mutation
in this gene is reported to associate with Donnai-Barrow and faciooculo-acoustico-renal syndromes15, and megalin knock-out mice
have high perinatal mortality owing to respiratory insufficiency16.
2
www.nature.com/scientificreports
Figure 1 | Signal plots and LD plots of two identified maternal TRD loci. (a) Regional signal plots of two identified maternal TRD loci. The TRD pvalues of the local SNPs (within 100 kb from the identified SNP) were transformed by a negative logarithm (to base 10) and then plotted on the
corresponding genomic positions (in Mb, NCBI build 36). The paternal and maternal TRD p-values are indicated by blue circles and red diamonds,
respectively; and the enlarged diamonds in the middle denote the identified SNPs. The genes within the local region and their directions are shown below
the dots. The red boxes on the position axes indicate LD regions in which the D’ values between most of the local SNPs and the identified SNPs were larger
than 0.5. (b) The LD plots of some local SNPs in the regions shown in (A) with red boxes. The LD blocks were inferred using Haploview according to the
D’ values indicated in the cells.
The second locus, rs926716, was located near the transcriptional
start site of ZNF133, and the maternal TRD signal extended towards
the upstream of the gene (Fig. 1). In contrast to the previous locus,
the preferred maternal transmitted allele of rs926716 was the minor
allele (Table 1). The TRD effect was comparable to that of the previous locus (minor allele transmission percentage 62.8%). ZNF133 is
a transcription factor and it is highly expressed in CD341 cells
(BioGPS). The exact function of this gene is not clear to date.
Discussion
After eliminating the great majority of potential incidences of TRD as
most likely being genotyping errors, we identify two SNPs showing
sex-specific (both maternal) TRD. A previous study also used the
FHS data to search for human TRD loci of SNPs, but without separately considered for paternal and maternal TRD7. As a result, thousands of false-positive TRD loci due to genotyping error were
identified and it was hard to reveal true TRD loci. In contrast, separate tests for paternal and maternal TRD can maintain the maximum
power to detect paternal/maternal-specific TRD loci, and it can also
help distinguish false-positive TRD loci because genotyping errors
should affect both paternal and maternal transmission ratios equally.
Nevertheless, the genotyping quality of the identified TRD loci is not
ideal, especially for rs6733122, and genotyping validation of the SNP
array data and replication in further samples from the Framingham
population are still necessary to confirm the TRD.
A recent study also used the FHS data to detect general and sexspecific TRD loci in humans17. Excluding those probably caused by
genotyping errors, no genome-wide-significant sex-specific TRD
SCIENTIFIC REPORTS | 3 : 2147 | DOI: 10.1038/srep02147
locus was identified. The two maternal TRD loci here were not identified in that study, probably because the transmissions of the
ambiguous trios were included in that study for paternal or maternal
transmission with a number of 0.5 for each trio as performed by
PLINK, which biased the test statistic and reduced the statistical
power. It has been proved that the appropriate way is to exclude
these ambiguous trios for tests of paternal or maternal TRD18.
An early sib-pair linkage study suggested that there was the trend
of extensive TRD throughout the human genome, and the trend was
contributed by many small-effect TRD loci, without evidence of sexdifferentiated TRD6. In this study, we showed in addition, that there
are also large-effect maternal-specific TRD loci but they are rare. The
TRD could be missed by linkage analysis because the markers used in
linkage studies may not well capture the TRD loci19.
The identified maternal-specific TRD loci may result from either
meiotic drive or postzygotic viability due to maternal effects (including disruptions to imprints). It is unknown what the mechanism of
action might be in either of the cases that we identified. Neither
marker is in a cytogenetic band with any connection to imprinting,
there being no predicted imprinted genes in either 20p11 or in
2q3120. Nevertheless, given that megalin knock-out mice show high
perinatal mortality it is worth speculating that the TRD of rs6733122
on LRP2 is caused by postzygotic viability due to a maternal effect16.
The TRD loci identified in this study are probably of functional
significance, and they may also relate to human diseases. Notably,
rs926716 is associated with blood cholesterol level, which is an intermediate phenotype related to many complex diseases. A previous
study found that a Crohn’s disease-associated allele on DLG5 is
3
www.nature.com/scientificreports
transmission-distorted among healthy male offspring21; and interestingly, a recent study found that a locus rs3792106 on ATG16L1,
which is also associated with Crohn’s disease, was subject to maternal
TRD in healthy controls22. We could not replicate the maternal TRD
of rs3792106 in the Framingham population (P 5 0.70), which may
be due to the fact that the TRD observed was population-specific.
In addition, we could not replicate the identified two maternal
TRD loci (rs6733122 and rs926716) using two other datasets,
namely, the HapMap Phase II data23 and the Autism Genetic
Resource Exchange data24 (all P . 0.05, data not shown). Rather
than declaring that the identified TRD might be statistical false positives, we speculate that large-effect TRD are likely to be observed in
some specific population rather than all populations, because largeeffect TRD loci are unstable, that is, the preferred alleles should
quickly dominate the population after the TRD started so it is extremely unlikely that one TRD can be observed in many populations
simultaneously.
Finally, it can be seen from this study that genome-wide-significant TRD loci could be identified in pedigrees without any enrichment of a certain disease in offspring, when the available sample size
is large enough. It is possible that, in a TDT with genome-wide SNP
data for a certain disease, there may be spurious TRD loci identified
that are not necessarily relevant to the specified disease. Therefore,
TDT results should be interpreted with caution, and an appropriate
control for the identified TRD loci, such as to sample and test a
number of unaffected trios from the studied population, would help
to avoid the situation mentioned above.
Methods
Framingham heart study data. The data analyzed here were obtained from the
GAW16 (Genetic Analysis Workshop 16) Problem 2: the Framingham Heart Study
data25, through dbGaP26 (Database of Genotypes and Phenotypes, study No.
phs000128, version 1). The data include ,7,000 phenotyped and genotyped
individuals from three-generation pedigrees in Framingham. The SNP genotype data
of Affymetrix GeneChip Human Mapping 500 K Array Set were used in this study.
These genotypes were obtained from raw intensities using the genotype calling
algorithm BRLMM of Affymetrix and the genotype data were directly available. No
genotyping batch information is available.
Quality control. The data mentioned above were processed with a series of qualitycontrol procedures, as follows. (1) Removal of any individual with missing sex
information; (2) removal of twins; (3) removal of any individual with a call rate ,
97%; (3) exclude any SNP with . 5% missing genotypes; (4) removal of any
individual with problematic sex information according to the sex inferred by the X
chromosome homozygosity estimate F (male: F . 0.8; female: F , 0.2); (5) removal of
out-of-pedigree singletons; (6) removal of potential duplicated samples if the
estimated proportion of identical-by-decent allele sharing between any two
individuals was larger than 0.9 (none removed); (7) removal of any offspring with an
individual-wise Mendelian error rate . 0.01 (none removed); (8) exclude any SNP
with a SNP-wise Mendelian error rate . 0.01; (9) exclude any SNP for which the
proportion of "informative" (as in transmission disequilibrium test, but here we
excluded ambiguous trios with all heterozygous genotypes) paternal or maternal
transmissions was lower than 0.05; this is a similar control to that for minor allele
frequency, but it was more appropriate in the TRD analysis to avoid sparse data; (10)
remove individuals that were not in any complete parent-offspring trios. Procedures
(1–6) were performed with the PLINK software27 and others were performed with
customized programs. After these procedures, the data included 348,989 SNPs and
2,315 individuals from 319 pedigrees.
Statistical analysis of paternal and maternal TRDs. As each child corresponds to a
parental transmission event, a total of 1,241 parent-offspring trios (for 1,241 children)
were derived from the 319 pedigrees for the analyses of paternal and maternal TRDs.
For each SNP, the numbers of paternal and maternal allele transmissions were
respectively recorded if the transmissions were informative. Transmissions in
ambiguous trios with all heterozygous genotypes were not included. We used the
goodness-of-fit x2 test to determine if the paternal and maternal transmission ratio of
two alleles of a certain SNP deviated from the expected ratio of 151, and used
Bonferroni correction to adjust the p-values employing the total number of multiple
tests, including the separate tests for paternal and maternal TRDs. A standard x2 test
was used to evaluate the difference between the paternal and maternal transmission
ratio for any given SNP. The analyses were performed with customized programs and
the results of the two identified TRD loci were verified with the S.A.G.E. software28.
Cross validation of the TRD loci. For each identified TRD locus, ten-fold cross
validation was performed and the cross validation was repeated 100 times. In each
SCIENTIFIC REPORTS | 3 : 2147 | DOI: 10.1038/srep02147
cross validation, the pedigrees were randomly divided into ten groups, and every
group was labeled by turn as testing data while the other nine groups were combined
and labeled as training data. In each turn, the validation was considered consistent if
the direction of the TRD in the testing data was the same as that in the training data
(that is, the preferred transmitted alleles in the testing data and the training data were
the same).
Phenotypic association analyses. We performed population-based phenotypic
association analyses using unrelated individuals of the FHS data. Here the individuals
removed in quanlity-control procedure (10) were retained because the individuals
that are not in complete trios can still be used in population-based association
analyses. Founders were extracted and then 444 individuals were removed due to
relatedness (estimated proportion of identical-by-descent allele sharing larger than
0.125, using PLINK). A total of 1,330 individuals remained for association analyses.
Eight quantitative traits and the two disease statuses were analyzed using linear
regression and logistic regression, respectively, on the two identified SNPs for the
evaluation of allelic additive effects. All traits except height were log-transformed to
approximate normal distribution. Sex, age and smoking status were included as
covariates.
Family-based association analyses were also performed using the same data as in
the TRD analyses. We used QTDT29 to perform the association analyses of paternal
and maternal allele transmissions of the two identified TRD loci with the eight
quantitative traits. The same covariates as those mentioned above were included.
TDT tests with the two diseases were not valid due to general TRD; therefore, we used
x2 tests to test for differences in allele transmission ratio between those trios with
affected offspring and the other trios. Paternal and maternal allele transmission ratios
were analyzed separately.
1. Kusano, A., Staber, C. & Ganetzky, B. Nuclear mislocalization of enzymatically
active RanGAP causes segregation distortion in Drosophila. Dev Cell 1, 351–61
(2001).
2. Bauer, H., Willert, J., Koschorz, B. & Herrmann, B. G. The t complex-encoded
GTPase-activating protein Tagap1 acts as a transmission ratio distorter in mice.
Nat Genet 37, 969–73 (2005).
3. Phadnis, N. & Orr, H. A. A single gene causes both male sterility and segregation
distortion in Drosophila hybrids. Science 323, 376–9 (2009).
4. Greenwood, C. M. & Morgan, K. The impact of transmission-ratio distortion on
allele sharing in affected sibling pairs. Am J Hum Genet 66, 2001–4 (2000).
5. Pardo-Manuel de Villena, F. & Sapienza, C. Transmission ratio distortion in
offspring of heterozygous female carriers of Robertsonian translocations. Hum
Genet 108, 31–6 (2001).
6. Zollner, S. et al. Evidence for extensive transmission distortion in the human
genome. Am J Hum Genet 74, 62–72 (2004).
7. Paterson, A. D. et al. Transmission-ratio distortion in the Framingham Heart
Study. BMC Proc 3 Suppl 7, S51 (2009).
8. Dawber, T. R., Meadors, G. F. & Moore, F. E., Jr. Epidemiological approaches
to heart disease: the Framingham Study. Am J Public Health Nations Health 41,
279–81 (1951).
9. Feinleib, M., Kannel, W. B., Garrison, R. J., McNamara, P. M. & Castelli, W. P. The
Framingham Offspring Study. Design and preliminary data. Prev Med 4, 518–25
(1975).
10. Splansky, G. L. et al. The Third Generation Cohort of the National Heart, Lung,
and Blood Institute’s Framingham Heart Study: design, recruitment, and initial
examination. Am J Epidemiol 165, 1328–35 (2007).
11. Becker, K. G., Barnes, K. C., Bright, T. J. & Wang, S. A. The genetic association
database. Nat Genet 36, 431–2 (2004).
12. Hindorff, L. A. et al. Potential etiologic and functional implications of genomewide association loci for human diseases and traits. Proc Natl Acad Sci U S A 106,
9362–7 (2009).
13. Barrett, J. C., Fry, B., Maller, J. & Daly, M. J. Haploview: analysis and visualization
of LD and haplotype maps. Bioinformatics 21, 263–5 (2005).
14. Wu, C. et al. BioGPS: an extensible and customizable portal for querying and
organizing gene annotation resources. Genome Biol 10, R130 (2009).
15. Kantarci, S. et al. Mutations in LRP2, which encodes the multiligand receptor
megalin, cause Donnai-Barrow and facio-oculo-acoustico-renal syndromes. Nat
Genet 39, 957–9 (2007).
16. Willnow, T. E. et al. Defective forebrain development in mice lacking gp330/
megalin. Proc Natl Acad Sci U S A 93, 8460–4 (1996).
17. Meyer, W. K. et al. Evaluating the evidence for transmission distortion in human
pedigrees. Genetics 191, 215–32 (2012).
18. Sham, P. C., Zhao, J. H., Waldman, I. & Curtis, D. Should ambiguous trios for the
TDT be discarded? Ann Hum Genet 64, 575–6 (2000).
19. Paterson, A. D., Sun, L. & Liu, X. Q. Transmission ratio distortion in families from
the Framingham Heart Study. BMC Genet 4 Suppl 1, S48 (2003).
20. Luedi, P. P. et al. Computational and experimental identification of novel human
imprinted genes. Genome Res 17, 1723–30 (2007).
21. Friedrichs, F. et al. Evidence of transmission ratio distortion of DLG5 R30Q
variant in general and implication of an association with Crohn disease in men.
Hum Genet 119, 305–11 (2006).
4
www.nature.com/scientificreports
22. Liu, L. Y., Schaub, M. A., Sirota, M. & Butte, A. J. Transmission distortion in
Crohn’s disease risk gene ATG16L1 leads to sex difference in disease association.
Inflamm Bowel Dis 18, 312–22 (2012).
23. Frazer, K. A. et al. A second generation human haplotype map of over 3.1 million
SNPs. Nature 449, 851–61 (2007).
24. Lajonchere, C. M. Changing the landscape of autism research: the autism genetic
resource exchange. Neuron 68, 187–91 (2010).
25. Cupples, L. A., Heard-Costa, N., Lee, M. & Atwood, L. D. Genetics Analysis
Workshop 16 Problem 2: the Framingham Heart Study data. BMC Proc 3 Suppl 7,
S3 (2009).
26. Mailman, M. D. et al. The NCBI dbGaP database of genotypes and phenotypes.
Nat Genet 39, 1181–6 (2007).
27. Purcell, S. et al. PLINK: a tool set for whole-genome association and populationbased linkage analyses. Am J Hum Genet 81, 559–75 (2007).
28. Elston, R. C. & Gray-McGuire, C. A review of the ‘Statistical Analysis for Genetic
Epidemiology’ (S. A. G. E.) software package. Hum Genomics 1, 456–9 (2004).
29. Abecasis, G. R., Cardon, L. R. & Cookson, W. O. A general test of association for
quantitative traits in nuclear families. Am J Hum Genet 66, 279–92 (2000).
Acknowledgements
We acknowledge the Genetic Analysis Workshop, the Framingham Heart Study, the
National Institute of Health, and their investigators who contributed to the data analyzed in
SCIENTIFIC REPORTS | 3 : 2147 | DOI: 10.1038/srep02147
this study. We also greatly thank the anonymous referees for their valuable comments and
suggestions which help improve the quality of the paper. This work is supported by the
National Basic Research Program of China (No. 2011CB510100), the National Natural
Science Foundation of China (No. 81030015), and the E-Institutes of Shanghai Municipal
Education Commission.
Author contributions
Y.L., X.K. and L.D.H. conceived and designed the experiments. Y.L. performed all the
statistical and computational analyses. Y.L. and L.D.H. wrote the paper. Y.L., L.Z., S.X., L.H.,
L.D.H., X.K. analyzed the results and reviewed the manuscript.
Additional information
Supplementary information accompanies this paper at http://www.nature.com/
scientificreports
Competing financial interests: The authors declare no competing financial interests.
How to cite this article: Liu, Y. et al. Identification of Two Maternal Transmission Ratio
Distortion Loci in Pedigrees of the Framingham Heart Study. Sci. Rep. 3, 2147;
DOI:10.1038/srep02147 (2013).
This work is licensed under a Creative Commons Attribution 3.0 Unported license.
To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0
5